We exist to make AI decisions defensible.

Mission

Virdant exists to make every AI decision explainable, traceable, and defensible. As AI agents take on higher-stakes roles — approving loans, flagging content, routing medical queries — the inability to explain a decision is not a minor inconvenience. It is a legal, ethical, and operational liability.

Origin

Virdant began as an internal tool built while working on AI-powered insurance claims processing. After the third time a compliance team asked "what exactly did the model evaluate, and why did it deny that claim?" — and the answer was "check the logs" — it became clear that logs are not audit records. They are evidence of activity, not evidence of reasoning. Virdant was built to close that gap.

Open source commitment

Virdant is MIT-licensed and will remain so. The SDK has zero runtime dependencies and stores records in standard JSONL. There is no hosted service, no locked-in format, no dependency on Virdant as an ongoing business. If this project disappeared tomorrow, every organization using it would still have their audit data in a format they can query with any tool that reads newline-delimited JSON.

That is the promise: your audit data is yours.